Skip to content

Tutorial 1: Using PyTechnicalIndicators with pandas

Leverage Rust-powered speed inside familiar pandas workflows.

This tutorial is the first in a series of tutorials:

🎯 Goal

In this tutorial you will learn how to:

  • Install and import pytechnicalindicators
  • Load OHLCV price data into a pandas DataFrame
  • Convert columns to the list inputs expected by the library
  • Call both single (scalar) and bulk (rolling / series) indicator functions
  • Add indicator outputs back into your DataFrame
  • Understand parameter and model choices (e.g. different moving average types)
  • Handle common pitfalls (NaNs, alignment, window offsets)

📦 Prerequisites

Requirement Notes
Python 3.10+ (Match your environment; verify with python --version)
pip / virtualenv Recommended to isolate dependencies
pandas For data handling

Install basics:

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install pandas

Install PyTechnicalIndicators (assuming published wheel / build locally):

pip install pytechnicalindicators
# OR (from source after cloning)
# pip install maturin
# maturin develop

🧱 Library Structure (Quick Recap)

Modules are grouped by analysis area:

  • standard_indicators
  • momentum_indicators
  • trend_indicators
  • strength_indicators
  • volatility_indicators
  • candle_indicators
  • chart_trends
  • correlation_indicators
  • other_indicators

Each module has two namespaces:

  • single: operates on the entire list (returns one value or a tuple)
  • bulk: rolling / windowed computation (returns a list aligned to the trailing end of each window)

Example:

from pytechnicalindicators import standard_indicators

standard_indicators.single.simple_moving_average(prices)
standard_indicators.bulk.simple_moving_average(prices, period=20)

📥 Getting Some Data

If you already have a CSV point to your CSV, we will be using examples/prices.csv for our tutorials:

df = pd.read_csv("prices.csv", parse_dates=["Date"])
df = df.sort_values("Date").reset_index(drop=True)

🔄 Converting DataFrame Columns to Lists

Most functions expect Python list[float] (not Series, not NumPy arrays). Convert cleanly:

close = df["Close"].astype(float).tolist()
high  = df["High"].astype(float).tolist()
low   = df["Low"].astype(float).tolist()
open_ = df["Open"].astype(float).tolist()

(You can keep naming consistent; open_ avoids shadowing Python's built-in open.)

✅ Example 1: Simple Moving Average (SMA)

API Reference

Single (all data as one window)

from pytechnicalindicators import import moving_average as ma

sma_all = ma.single.moving_average(close, moving_average_type="simple")
print("Full-series SMA:", sma_all)

Bulk (rolling window)

period = 20
sma_series = ma.bulk.moving_average(close, period=period, moving_average_type="simple")
# Align back to DataFrame: last len(sma_series) rows correspond to rolling results
df.loc[df.index[-len(sma_series):], f"SMA_{period}"] = sma_series

Tip: The first (period - 1) rows have no value because a full window was not yet available. You can forward fill or leave as NaN depending on downstream needs.

✅ Example 2: Moving Constant Bands (Generic Bollinger Bands)

API Reference

single version returns a tuple (lower, middle, upper); bulk returns a list of such tuples.

lower, middle, upper = ci.single.moving_constant_bands(
        close[-20:],
        constant_model_type="exponential_moving_average",
        deviation_model="standard_deviation",
        deviation_multiplier=2.0,
)
print(lower, middle, upper)

Bulk usage:

bands = ci.bulk.moving_constant_bands(
    close,
    constant_model_type="exponential_moving_average",
    deviation_model="standard_deviation",
    deviation_multiplier=2.0,
    period=20
)

# Unpack and assign (align to tail)
lower_vals, mid_vals, upper_vals = zip(*bands)
tail_index = df.index[-len(bands):]
df.loc[tail_index, "MCB_Lower"] = lower_vals
df.loc[tail_index, "MCB_EMA"] = mid_vals
df.loc[tail_index, "MCB_Upper"] = upper_vals

✅ Example 3: Relative Strength Index (RSI)

API Reference

rsi_values =  mi.bulk.relative_strength_index(
    close,
    constant_model_type="smoothed_moving_average",
    period=20
)

df.loc[df.index[-len(rsi_values):], "RSI"] = rsi_values

✅ Example 4: Average True Range (ATR)

API Reference

Valid constant model strings (case-insensitive aliases):
simple, smoothed, exponential, median, mode
(Full forms also accepted: simple_moving_average, smoothed_moving_average, etc.)

from pytechnicalindicators import other_indicators

model = "exponential_moving_average"
period = 14

atr_series = other_indicators.bulk.average_true_range(
    close=close,
    high=high,
    low=low,
    constant_model_type=model,
    period=period
)
df.loc[df.index[-len(atr_series):], f"ATR_{period}"] = atr_series

🧪 Single vs Bulk: When to Use Which?

Use Case Pick
One summary value for entire dataset single
Rolling feature engineering for models bulk
Signal snapshot at latest bar single on a sliced tail
Chart overlays / time series indicators bulk

🛠️ Handling Alignment

Bulk outputs are shorter than the original price history when a window (period) is specified. Strategies:

# Method 1: Right-align (as shown earlier)
df.loc[df.index[-len(series):], "Indicator"] = series

# Method 2: Pad with NaN at front
import numpy as np
padded = [np.nan] * (len(df) - len(series)) + series
df["Indicator"] = padded

🧪 Missing Data / NaNs

Before extracting lists:

df = df.dropna(subset=["Open", "High", "Low", "Close", "Volume"])

If you forward-fill:

df[["Open","High","Low","Close","Volume"]] = df[["Open","High","Low","Close","Volume"]].ffill()

Be cautious: forward-filling can create artificial continuity.

⚙️ Choosing Constant / Deviation Models

Many functions accept a constant_model_type or deviation_model to alter smoothing or dispersion behavior.
Experiment:

Consult the project API reference for model-specific effects:

🧩 Putting It All Together

A runnable example of the full code can be found in 01_using_pandas_and_pytechnicalindicators.py

python3 01_using_pandas_and_pytechnicalindicators.py

Output:

Full-series SMA: 5624.867410358566
5367.04826935573 5755.587345685254 6144.126422014779
          Date     Open     High      Low    Close     SMA_20     BB_Lower    BB_Middle     BB_Upper        RSI   ATR_20
246 2025-03-10  5705.37  5705.37  5564.02  5614.56  5956.2700  5579.906354  5877.649008  6175.391662  33.281207  84.9520
247 2025-03-11  5603.65  5636.30  5528.41  5572.07  5931.5515  5504.588851  5841.191158  6177.793465  36.744063  88.9195
248 2025-03-12  5624.84  5642.19  5546.09  5599.30  5908.0915  5451.413841  5811.173289  6170.932737  33.949163  92.0275
249 2025-03-13  5594.45  5597.78  5504.65  5521.52  5881.5690  5385.361698  5775.695387  6166.029075  33.043555  93.6840
250 2025-03-14  5563.85  5645.27  5563.85  5638.94  5857.7625  5367.048269  5755.587346  6144.126422  39.573869  94.4570

🚀 Performance Tips

  • Convert Series to plain Python lists only once; reuse.
  • Gather all indicators first, then assign to DataFrame to minimize index alignment overhead.
  • For large backtests, consider chunking historical data and streaming features forward.

🧾 Common Pitfalls

Issue Cause Fix
ValueError for model type Misspelled string Use accepted aliases (simple, exponential, etc.)
Mismatched lengths when merging Rolling window shrinks output Right-align or pad with NaNs
NaNs propagate into indicators Missing OHLC data Clean with dropna or controlled fills
Confusing single vs bulk Wrong namespace Use module.single.* or module.bulk.*

🛡️ Disclaimer

Educational example only. Not financial advice. Validate results independently before using in production or live trading.

✅ Next Step

You now know how to integrate PyTechnicalIndicators into a pandas workflow and engineer feature columns efficiently.

Next discover how to integrate Plotly charts

Happy analyzing! 🦀🐍📈